CNN-RNN_Project

Fundamentals of CNN and RNN with keras & tensorflow libs

Objectives

CNN part

• We are to implement a multi-class image classifier using CNNs

• Original images of 224 * 224 pixel, 15 different classes (of vegetables)

• We should consider 2 input sizes (grayscale images):

  • 31 * 31 Pixels - labelled "a" throughout this notebook
  • 128 * 128 Pixels - labelled "b" throughout this notebook

RNN part

• Building a next word predictor, given a sequence of words
• We would try to predict the next 10 words for each given input of words
• Find ways to evaluate generated sequence of words (how meaningful / creative), using seed_texts as benchmark